Hi Joshua,
A script might get you part of the way there, but I don't think you'll get everything you want out of it.
E.g. you could change the bid modifiers in a campaign, then schedule a script which would reset the modifiers and report on stats from the experimental period vs. the previous time period. But testing something like that over two different time periods introduces a bunch of confounding factors; a proper experiment would run concurrently with the normal bid modifier, and could report stats for the same time period.
Your best bet here might be to just make a copy of the campaign, share a budget between that and the original, and then see which campaign performs better. But that approach also has its share of problems.
Sorry I couldn't be of more help here, but this is just a limitation of how experiments work -- they affect entire entities (keywords, ads, etc.) instead of affecting each property (bids, bid modifiers, targeting, etc.).